Beyond Context7: Self-Host Your Own Documentation MCP Server

If you're using AI coding assistants like Claude Code, Cursor, or Windsurf, you've likely used Context7 — the MCP server that fetches documentation for thousands of public libraries. Type a question about React, Next.js, or Prisma, and Context7 retrieves the latest docs straight from its indexed catalog. It's a game-changer for AI-assisted development.

But Context7 has a blind spot: your code.

Your private repositories. Your internal microservices. Your team's bespoke framework. The 200,000-line monorepo with zero documentation that you just inherited. Context7 can't index these because they're not in its public catalog — and they shouldn't be.

That's where ContextMCP Dockerized comes in.

Context7: What It Does Well

Context7 is a hosted service that indexes documentation from thousands of open-source libraries. When your AI assistant needs to know the latest Supabase API or how to use a specific Tailwind utility, Context7 delivers the answer with source-grounded documentation.

It's a powerful augmentation, and we use it every day. But it's not designed for:

• Private/internal documentation — Your team's architecture docs, ADRs, or runbooks
• Proprietary codebases — Your company's monorepo or internal libraries
• Niche or unindexed libraries — Frameworks not in Context7's catalog
• Local development docs — READMEs, API specs, and conventions in your local repos

ContextMCP Dockerized: The Alternative

ContextMCP Dockerized wraps the upstream ContextMCP project in a production-ready Docker setup, making it trivial to run your own documentation MCP server — fully local, fully private, and indexing exactly what you need.

Local Mode (Full Privacy)

Everything runs on your machine:

• Ollama for local embeddings (nomic-embed-text, no external API calls)
• Pinecone Local for in-memory vector storage (no cloud dependency)
• Your docs — indexed from local repos or GitHub

No data ever leaves your machine. No API keys required. Zero ongoing costs.

Cloud Mode (Scale)

When you need more power or shared team access:

• OpenAI embeddings for higher-quality vectors
• Pinecone Cloud for persistent, scalable vector search
• Same configuration — switch between local and cloud with a single flag

Head-to-Head: Context7 vs. ContextMCP Dockerized

Better Together: Augment, Don't Replace

Context7 and ContextMCP Dockerized are complementary, not mutually exclusive. The best AI-assisted development setup uses both:

Run both MCP servers side by side. Your AI assistant queries Context7 for library references and your local ContextMCP server for project-specific knowledge. No overlap, no conflict — just complete coverage.

Multi-Repo Indexing: The Killer Feature

One of the most powerful features of ContextMCP Dockerized is multi-repo indexing. Instead of configuring one repository at a time, you can point it at a directory of projects:

Each repo is indexed with AST-aware parsing — extracting methods, classes, and functions from code files while treating config files as whole documents. The result is a semantic search index that understands your codebase structure, not just text.

Getting Started in 5 Minutes

1. Clone the repo: git clone https://github.com/J3ys/ContextMCPDocker
2. Configure: Edit config.yaml with your repository
3. Start local infra: docker compose -f docker-compose.local.yml up -d pinecone ollama
4. Index your docs: docker compose -f docker-compose.local.yml --profile reindex run --rm reindex
5. Run the app: docker compose -f docker-compose.local.yml up -d app

That's it. Your MCP server is running and ready to answer questions about your codebase. Add it as an MCP server in your AI coding tool and start querying your own documentation.

What's Coming: A Dedicated Fork

ContextMCP Dockerized currently builds on top of the upstream dodopayments/context-mcp project with a small local-support patch. But we're actively working on a dedicated fork that will:

• Add native multi-repo support in the core engine
• Improve the reindex pipeline for incremental updates
• Support more embedding providers and vector stores
• Provide a web UI for managing indexed sources

The fork will remain fully open source — because documentation access shouldn't be locked behind a paywall.

Open Source: Built by Sendel.org

ContextMCP Dockerized is developed by Sendel.org (that's me), an independent developer focused on AI orchestration, enterprise architecture, and e-commerce systems. The project is fully open source under the original ContextMCP license.

We need your help. This is early-stage software, and there's a lot to build:

• Better documentation and examples
• Support for more vector databases
• Improved chunking strategies for different codebases
• A web UI for configuration and indexing
• CI/CD integration for automatic reindexing

Whether you're fixing a bug, improving the docs, or adding a feature — pull requests are warmly welcomed.

Conclusion: You Deserve Both

Context7 is excellent at what it does — providing instant access to public library documentation. It's a tool we use and recommend.

But if you're building software, most of the documentation you need is your own. Your internal APIs, your team conventions, your architecture decisions. That knowledge shouldn't be locked in people's heads or scattered across wiki pages.

ContextMCP Dockerized brings that knowledge into your AI assistant's reach. Run it locally for privacy, run it in the cloud for team access, and soon — run it on a dedicated fork with even more features.

Your AI assistant is only as good as the context you give it. Give it everything.